We use a custom imaging robot to obtain the time-lapse images of the individual experiments. This is a simple but effective build. Two stepper motors control the x and y axes of motion. A camera mount takes pictures of each experiment as the robot moves across the shelf. Each transparent container holds one experiment of a certain rice variety. The containers are identified through a QR code label. Typically, one batch of experiments will stay on the shelf for 2-3 weeks. The robot will then upload the images for processing.
Here is a close look at two time-lapse videos that have been spliced together. S1 rice seed type (left) vs S1 mutant type (right)
Changing the substrate yields interesting results. Here are rice seeds growing in gravel. The gel concentration of the primary transparent substrate can also be tuned.
Qualitative analysis is done after the initial time-lapse videos have been uploaded. One of the most insightful data points we can collect are the root coordinates by frame. This is done with the help of object detection. We have trained a neural network to recognize the root tip in a single frame. The post-processing pipeline has a number of steps in order sort, analyze, and archive the incoming experiments.
Research Supervisor
Software Developer